Table 2.

Classification performance of the ¹²³I-FP-CIT SPECT SVM analysis

	Sp (%)	Sn (%)	Acc (%)	AUC	Highest beta weights	p value
Binary classification of PS vs CTL
All PS vs CTL	96.7	89.5	92.9	0.97	S (more) C/P ratio (more) S-AI	< 0.001
Binary classifications of each combination of two PS
PD vs MSAP	95 (fixed)	79.4		0.66	C (mean) C/P ratio (less) P-AI	< 0.001
	63.8	58.2	61.1
PD vs PSP	95 (fixed)	86		0.71	C (more) C (less) P-AI	< 0.001
	61.9	66.9	64.2
PD vs CBS	95 (fixed)	82.3		0.78	C/P ratio (more) C/P ratio (less) S-AI	< 0.001
	76.2	65.9	72.0
MSAP vs PSP	64	64	62.9	0.64	S (less) C/P ratio (more) C-AI	0.07
MSAP vs CBS	84.6	82.6	83.7	0.95	P (mean) P (less) S-AI	< 0.001
PSP vs CBS	72.7	82.6	73.9	0.79	P (less) S (mean) S-AI	< 0.001
Binary classifications of each PS vs all other PS
PD vs non-PD	69.9	45	58.4	0.60	C (more) C/P ratio (less) S-AI	0.01
MSA-P vs non-MSA-P	77.1	71.4	74.4	0.78	C/P ratio (more) C-AI	0.002
PSP vs non-PSP	54.2	66.7	60	0.58	C/P ratio (more) S-AI	0.06
CBS vs non-CBS	83.3	69.1	76.7	0.88	P (mean) S-AI	< 0.001